Cover Image

Toward universal steering and monitoring of AI models

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Beaglehole, Daniel, Radhakrishnan, Adityanarayanan, Boix-Adserà, Enric, Belkin, Mikhail
Format:	Preprint
Published:	2025
Subjects:	Computation and Language Artificial Intelligence Machine Learning
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Towards a theory of model distillation
by: Boix-Adsera, Enric
Published: (2024)

Secret mixtures of experts inside your LLM
by: Boix-Adsera, Enric
Published: (2025)

On the inductive bias of infinite-depth ResNets and the bottleneck rank
by: Boix-Adsera, Enric
Published: (2025)

The Features at Convergence Theorem: a first-principles alternative to the Neural Feature Ansatz for how networks learn representations
by: Boix-Adsera, Enric, et al.
Published: (2025)

xRFM: Accurate, scalable, and interpretable feature learning models for tabular data
by: Beaglehole, Daniel, et al.
Published: (2025)

Let Me Think! A Long Chain-of-Thought Can Be Worth Exponentially Many Short Ones
by: Mirtaheri, Parsa, et al.
Published: (2025)

The Weight Gram Matrix Captures Sequential Feature Linearization in Deep Networks
by: Cha, Taehun, et al.
Published: (2026)

The power of fine-grained experts: Granularity boosts expressivity in Mixture of Experts
by: Boix-Adsera, Enric, et al.
Published: (2025)

Contextual Linear Activation Steering of Language Models
by: Hsu, Brandon, et al.
Published: (2026)

When can transformers reason with abstract symbols?
by: Boix-Adsera, Enric, et al.
Published: (2023)

Emergence in non-neural models: grokking modular arithmetic via average gradient outer product
by: Mallinar, Neil, et al.
Published: (2024)

Catching rationalization in the act: detecting motivated reasoning before and after CoT via activation probing
by: Mirtaheri, Parsa, et al.
Published: (2026)

Linear Recursive Feature Machines provably recover low-rank matrices
by: Radhakrishnan, Adityanarayanan, et al.
Published: (2024)

Convergent Evolution: How Different Language Models Learn Similar Number Representations
by: Fu, Deqing, et al.
Published: (2026)

Efficient semantic uncertainty quantification in language models via diversity-steered sampling
by: Park, Ji Won, et al.
Published: (2025)

Quadratic models for understanding catapult dynamics of neural networks
by: Zhu, Libin, et al.
Published: (2022)

Can sparse autoencoders be used to decompose and interpret steering vectors?
by: Mayne, Harry, et al.
Published: (2024)

Context-Scaling versus Task-Scaling in In-Context Learning
by: Abedsoltan, Amirhesam, et al.
Published: (2024)

Catapults in SGD: spikes in the training loss and their impact on generalization through feature learning
by: Zhu, Libin, et al.
Published: (2023)

Towards Conversational Diagnostic AI
by: Tu, Tao, et al.
Published: (2024)

Efficient and accurate steering of Large Language Models through attention-guided feature learning
by: Davarmanesh, Parmida, et al.
Published: (2026)

Merge to Mix: Mixing Datasets via Model Merging
by: Tao, Zhixu Silvia, et al.
Published: (2025)

Domain-level metacognitive monitoring in frontier LLMs: A 33-model atlas
by: Cacioli, Jon-Paul
Published: (2026)

Towards Conversational AI for Disease Management
by: Palepu, Anil, et al.
Published: (2025)

Towards Execution-Grounded Automated AI Research
by: Si, Chenglei, et al.
Published: (2026)

ClimateGPT: Towards AI Synthesizing Interdisciplinary Research on Climate Change
by: Thulke, David, et al.
Published: (2024)

The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery
by: Lu, Chris, et al.
Published: (2024)

Large Language Models in the Task of Automatic Validation of Text Classifier Predictions
by: Tsymbalov, Aleksandr, et al.
Published: (2025)

RaguTeam at SemEval-2026 Task 8: Meno and Friends in a Judge-Orchestrated LLM Ensemble for Faithful Multi-Turn Response Generation
by: Bondarenko, Ivan, et al.
Published: (2026)

Towards Automated Patent Workflows: AI-Orchestrated Multi-Agent Framework for Intellectual Property Management and Analysis
by: Srinivas, Sakhinana Sagar, et al.
Published: (2024)

AI-AI Bias: large language models favor communications generated by large language models
by: Laurito, Walter, et al.
Published: (2024)

The merged-staircase property: a necessary and nearly sufficient condition for SGD learning of sparse functions on two-layer neural networks
by: Abbe, Emmanuel, et al.
Published: (2022)

Towards AI-$45^{\circ}$ Law: A Roadmap to Trustworthy AGI
by: Yang, Chao, et al.
Published: (2024)

Feature learning as alignment: a structural property of gradient descent in non-linear neural networks
by: Beaglehole, Daniel, et al.
Published: (2024)

Out-of-Distribution Detection using Synthetic Data Generation
by: Abbas, Momin, et al.
Published: (2025)

Wikontic: Constructing Wikidata-Aligned, Ontology-Aware Knowledge Graphs with Large Language Models
by: Chepurova, Alla, et al.
Published: (2025)

Limitations of Normalization in Attention Mechanism
by: Mudarisov, Timur, et al.
Published: (2025)

Scaling Transformer to 1M tokens and beyond with RMT
by: Bulatov, Aydar, et al.
Published: (2023)

AI and Generative AI for Research Discovery and Summarization
by: Glickman, Mark, et al.
Published: (2024)

BEExAI: Benchmark to Evaluate Explainable AI
by: Sithakoul, Samuel, et al.
Published: (2024)